Build with Mellix
Ask a question about a situation and get a calibrated probability for every option, through our hosted API or on hardware you own.
Mellix builds decision models, starting with the Myles family. You describe a situation, ask a question with the options you care about, and the model gives every option a probability in one pass. Your code acts on the confident answers and hands the rest to a person. The models decide; they do not write text.
How a decision works
- Describe the situation. A message, a record, a policy and a case, as text or JSON. Myles-G can read images too.
- Ask a typed question. Yes or no (
noul), one of several labels (choice) or a level on a scale (score). See Question types. - Read the probabilities. Every option gets one. Act on what clears your threshold and send the rest to a person; Thresholds shows how.
Request
- model
- "myles-g-12b"
- state
- "Your bank account has been locked…"
- question
- "Is this message a scam?"
- type
- "noul"
Response
- Yes0.97
- No0.03
What you can decide
- Is this message a scam?
- Which team should take it?
- Does this contain personal details?
- Is this answer supported by its source?
Every decision is a question with options. Use cases lists sixteen capabilities across twelve industries.
Choose a model
| Where it runs | Model |
|---|---|
| A browser or a phone | Myles-22M |
| A laptop | Myles-150M |
| The cheapest server you have | Myles-400M |
| One GPU you own, or our hosted API | Myles-G 12B and Myles-G 31B |
| Laptop- and phone-class GPUs, by request | Myles-G Edge |
Models compares them, and Model types shows how each is called. Requests to the hosted API are encrypted with HTTPS and need your API key; Authentication and Data cover keys and what is stored.

